INDEX

Explanations

elements and attributes associated with HTML and web page structure

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L5_Hcat_z_lr1.20e-03_l11.00e+00_ds49152_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

float32

Hook Name

blocks.5.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.03

1:0.02

2:0.16

3:0.11

4:0.06

5:0.05

6:0.05

7:0.08

8:0.05

9:0.07

10:0.16

11:0.10

Negative Logits

 diluted

-1.30

Phi

-1.25

 accountable

-1.24

CoC

-1.23

 Sigma

-1.22

 pharmac

-1.21

 commissions

-1.18

ACS

-1.17

UV

-1.16

 criminally

-1.16

POSITIVE LOGITS

>:

1.40

>(

1.40

+=

1.39

EGIN

1.33

":[

1.31

':

1.30

":-

1.22

='

1.20

1.19

Object

1.18

Activations Density 0.007%

elements and attributes associated with HTML and web page structure

No Known Activations